[systemd-devel] [PATCH] resolved: re-add support for getting local domain from DHCP

[Zbigniew Jędrzejewski-Szmek]

On Mon, Aug 04, 2014 at 05:21:46PM +0200, Lennart Poettering wrote:
> On Wed, 30.07.14 00:37, Zbigniew Jędrzejewski-Szmek (zbyszek at in.waw.pl) wrote:
> 
> > > +_public_ int sd_network_get_domainname(int ifindex, char **domainname) {
> > > +        _cleanup_free_ char *s = NULL, *p = NULL;
> > > +        int r;
> > > +
> > > +        assert_return(ifindex > 0, -EINVAL);
> > > +        assert_return(domainname, -EINVAL);
> > > +
> > > +        if (asprintf(&p, "/run/systemd/netif/links/%d", ifindex) < 0)
> > > +                return -ENOMEM;
> > Not terribly important, but please spell that as:
> > 
> >            char p[sizeof("/run/systemd/netif/links/") + DECIMAL_STRING_MAX(int)];
> > 	   snprintf(p, sizeof(p), "/run/systemd/netif/links/%d",
> > 	   ifindex);
> 
> Actually, I'd even use normal sprintf() here, not snprintf(), after all,
> we carefully sized the string anyway, hence no need to enforce the size
> limit here. Using sprintf() here makes is
> clear to me that the buffer was carefully sized before, so I think would
> be preferrable...
On my TODO list is adding snprintf_check (or snprintf_assert, or snprintf_sized,
not sure about the name), which would wrap the snprintf in an assert
to check that there was no truncation.

> In particular since snprintf() doesn't add a trailing NULL
> if it truncates the string, which makes the whole excercise kinda
> pointless... an snprintf() without something like char_array_0() invoked
> right after it always raises my eyebrows. 
Are you sure?

snprintf(3) implies that the terminating byte is always written. Testing
confirms that fwiw.

Zbyszek